Derivation and evaluation of baseline creatinine equations for hospitalized children and adolescents: the AKI baseline creatinine equation

Kidney and blood pressure outcomes 11 years after pediatric critical illness and longitudinal impact of AKI: a prospective cohort study

BACKGROUND

Acute kidney injury (AKI) is common in critically ill children and associated with adverse short-term outcomes; however, long-term outcomes are not well described.

METHODS

This longitudinal prospective cohort study examined the prevalence of chronic kidney disease (CKD) and hypertension (HTN) 11 vs. 6 years after pediatric intensive care unit (PICU) admission and association with AKI. We examined children (age < 19 years) without pre-existing kidney disease 11 ± 1.5 years after PICU admission at a single center. AKI was defined using serum creatinine criteria. The primary outcome was a composite of CKD or HTN. CKD was defined as estimated glomerular filtration rate (eGFR) < 90 mL/min/1.73 m2 or albuminuria. Multivariable analyses compared outcomes at 11- vs. 6-year follow-up and association with AKI during PICU admission.

RESULTS

Of 96 children evaluated 11 years after PICU admission, 16% had evidence of CKD or HTN (vs. 28% at 6 years, p < 0.05). Multivariable analysis did not show improvement in outcomes from 6- to 11-year follow-up. eGFR decreased from 6- to 11-year follow-up (adjusted coefficient - 11.7, 95% CI - 17.6 to - 5.9) and systolic and diastolic blood pressures improved. AKI was associated with composite outcome at 6-year (adjusted odds ratio (aOR) 12.7, 95% CI 3.2-51.2, p < 0.001), but not 11-year follow-up (p = 0.31). AKI was associated with CKD (aOR 10.4, 95% CI 3.1-34.7) at 11 years.

CONCLUSIONS

This study provides novel data showing that adverse kidney and blood pressure outcomes remain highly prevalent 10 years after critical illness in childhood. The association with AKI wanes over time.

Identifying acute kidney injury in children: comparing electronic alerts with health record data

BACKGROUND

Electronic (e-)alerts for rising serum creatinine values are increasingly used as clinical indicators of acute kidney injury (AKI). The aim of this study was to investigate to what degree AKI episodes, as identified using e-alerts, correlated with coding for AKI in the hospital record for a national cohort of hospitalised children and examine whether coding corresponded with 30-day mortality after an AKI episode.

METHODS

A cross-section of AKI episodes based on alerts issued for children under 18 years in England during 2017 were linked to hospital records. Multivariable logistic regression was used to examine patient and clinical factors associated with AKI coding. Agreement between coding and 30-day mortality was examined at hospital level.

RESULTS

6272 AKI episodes in 5582 hospitalised children were analysed. Overall, coding was poor (19.7%). Older age, living in the least deprived quintile (odds ratio (OR) 1.4, 95% Confidence Interval (CI) 1.1, 1.7) and higher peak AKI stage (stage 1 reference; stage 2 OR 2.0, 95% CI 1.7, 2.4; stage 3 OR 8.6, 95% CI 7.1, 10.6) were associated with higher likelihood of coding in the hospital record. AKI episodes during birth admissions were less likely to be coded (OR 0.4, 95% CI 0.3, 0.5). No correlation was seen between coding and 30-day mortality.

CONCLUSIONS

The proportion of AKI alert-identified episodes coded in the hospital record is low, suggesting under-recognition and underestimation of AKI incidence. Understanding the reasons for inequalities in coding, variation in coding between hospitals and how alerts can enhance clinical recognition is needed.

Estimating baseline creatinine levels based on the kidney parenchymal volume

BACKGROUND

Acute kidney injury (AKI) diagnosis often lacks a baseline serum creatinine (Cr) value. Our study aimed to create a regression equation linking kidney morphology to function in kidney donors and chronic kidney disease patients. We also sought to estimate baseline Cr in minimal change disease (MCD) patients, a common AKI-predisposing condition.

METHODS

We analyzed 119 participants (mean age 60 years, 50% male, 40% donors) with CT scans, dividing them into derivation and validation groups. An equation based on kidney parenchymal volume (PV) was developed in the derivation group and validated in the validation group. We estimated baseline Cr in 43 MCD patients (mean age 45 years, 61% male) using the PV-based equation and compared with their 6 month post-MCD onset Cr values.

RESULTS

In the derivation group, the equation for the estimated glomerular filtration rate (eGFR) was: eGFR (mL/min/1.73m2) = 0.375 × PV (cm3) + (- 0.395) × age (years) + (- 2.93) × male sex + (- 13.3) × hypertension + (- 14.0) × diabetes + (- 0.210) × height (cm) + 82.0 (intercept). In the validation group, the eGFR and estimated Cr values correlated well with the measured values (r = 0.46, p = 0.01; r = 0.51, p = 0.004, respectively). In the MCD group, the baseline Cr values were significantly correlated with the estimated baseline Cr values (r = 0.52, p < 0.001), effectively diagnosing AKI (kappa = 0.76, p < 0.001).

CONCLUSIONS

The PV-based regression equation established in this study holds promise for estimating baseline Cr values and diagnosing AKI in patients with MCD. Further validation in diverse AKI populations is warranted.

Artificial Intelligence in Pediatric Nephrology-A Call for Action

Artificial intelligence is playing an increasingly important role in many fields of clinical care to assist health care providers in patient management. In adult-focused nephrology, artificial intelligence is beginning to be used to improve clinical care, hemodialysis prescriptions, and follow-up of transplant recipients. This article provides an overview of medical artificial intelligence applications relevant to pediatric nephrology. We describe the core concepts of artificial intelligence and machine learning and cover the basics of neural networks and deep learning. We also discuss some examples for clinical applications of artificial intelligence in pediatric nephrology, including neonatal kidney function, early recognition of acute kidney injury, renally cleared drug dosing, intrapatient variability, urinary tract infection workup in infancy, and longitudinal disease progression. Furthermore, we consider the future of artificial intelligence in clinical pediatric nephrology and its potential impact on medical practice and address the ethical issues artificial intelligence raises in terms of clinical decision-making, health care provider-patient relationship, patient privacy, and data collection. This article also represents a call for action involving those of us striving to provide optimal services for children, adolescents, and young adults with chronic conditions.

Diagnostic Performance of the Acute Kidney Injury Baseline Creatinine Equations in Children and Adolescents with Type 1 Diabetes Mellitus Onset

Three new equations for calculating the estimated basal serum creatinine (ebSCr) in hospitalized children have been developed: the simplified acute kidney injury (AKI) baseline creatinine (ABC) equation which considered only age in the formula; the equation including age and minimum creatinine (Crmin) within the initial 72 h from hospitalization (ABC-cr); and the equation including Crmin and height, weight, and age as squared values (ABC-advanced). We aimed to test the diagnostic performance of the ABC, ABC-cr and ABC-advanced equations in diagnosing AKI in 163 prospectively enrolled children with type 1 diabetes mellitus (T1DM) onset. We considered measured basal serum creatinine (mbSCr), the creatinine measured 14 days after T1DM onset. AKI was defined by the highest/basal serum creatine (HC/BC) ratio > 1.5. On the basis of the mbSCr, the AKI was diagnosed in 66/163 (40.5%) patients. This prevalence was lower than the prevalence of AKI diagnosed on the basis of ABC ebSCr (122/163 patients; 74.8%) (p < 0.001) and similar to the prevalence of AKI diagnosed on the basis of ABC-cr ebSCr (72/163 patients; 44.2%) (p = 0.5) and to the prevalence of AKI diagnosed on the basis of ABC-advanced ebSCr (69/163; 42.3%) (p = 0.73). AKI determined using ABC ebSCr, ABC-cr ebSCr and ABC-advanced ebSCr showed, respectively, 63.5% (kappa = 0.35; p < 0.001), 87.7% (kappa = 0.75; p < 0.001), and 87.1% (kappa = 0.74; p < 0.001) agreement with AKI determined using mbSCr. Using the HC/BC ratio calculated on the basis of mbSCr as gold standard, for Bland−Altman plots the HC/BC ratio calculated on the basis of ABC formula presented higher bias and wider limits of agreement compared with the HC/BC ratio calculated on the basis of ABC-cr and ABC-advanced formulas. In the receiver−operating characteristics (ROC) curve analysis the HC/BC ratio calculated on the basis of ABC ebSCr presented lower area under the ROC curve (AUROC) (AUROC = 0.89; 95%CI: 0.85−0.95; p < 0.001) compared with HC/BC ratio calculated on the basis of ABC-cr (AUROC = 0.94; 95%CI: 0.91−0.98; p < 0.001) or ABC-advanced ebSCr (AUROC = 0.914; 95%CI: 0.91−0.97; p < 0.001). In both Bland−Altman plots and ROC curve analysis, the ABC-cr and ABC-advanced formulas performed similarly. In conclusion, the ABC-cr and ABC-advanced formulas present very good diagnostic performance toward AKI identification in a population of children with T1DM onset.